A typical result in Numerical Analysis is an algorithm to solve a mathematical problem. The main ingredients are the precise statement of the problem, an algorithm to solve it, and an assessment of the errors. From algorithm, there is a long journey to a program: the method needs to be more user-friendly (handling of error cases, etc.), has to be implemented in a programming language, and tested. The resulting program is not only a practical product, but also a very precise description of the algorithm. If the reader is a computer, nothing is left to interpretation as long as the translator is correct. In my talk, I would like to introduce an effort to preserve mathematical software: the histoRicalg project, and summarize some of its results.
In two decades from 1970 to 1990, there was an upsurge in the writing and publication of mathematical software. Much of this has been collected in subroutine libraries or published in print or in machine readable form. Traditionally, scientific results are published in journals, and journal papers are archived in libraries. However, as programs grew larger, they were no longer printed with the papers, but sent to interested readers on tape, diskettes, or other forms of data storage.
The knowledge accumulated in these codes is extensive and valuable. Its preservation, however, is more difficult than that of papers. Storage technologies come and go, programming languages come into fashion, then become forgotten. To preserve the algorithms, one needs to constantly copy their source code from old computers to newer ones and preserve the tools needed to compile them.
In this poster, I plan to introduce one effort to preserve mathematical codes: the histoRicalg project. Its primary objective is to find, document, and preserve the originals of the numerical codes that have found their way into the R statistical programming system. In particular, I will summarize my experience with attempts to run the originals of these codes on a modern Linux PC.
A typical result in Numerical Analysis is an algorithm to solve a mathematical problem. The main ingredients are the precise statement of the problem, an algorithm to solve it, and an assessment of the errors. From algorithm, there is a long journey to a program: the method needs to be more user-friendly (handling of error cases, etc.), has to be implemented in a programming language, and tested. The resulting program is not only a practical product, but also a very precise description of the algorithm. If the reader is a computer, nothing is left to interpretation as long as the translator is correct.
In two decades from 1970 to 1990, there was an upsurge in the writing and publication of mathematical software. Much of this has been collected in subroutine libraries or published in print or in machine readable form. Traditionally, scientific results are published in journals, and journal papers are archived in libraries. However, as programs grew larger, they were no longer printed with the papers, but sent to interested readers on tape, diskettes, or other forms of data storage.
The knowledge accumulated in these codes is extensive and valuable. Its preservation, however, is more difficult than that of papers. Storage technologies come and go, programming languages come into fashion, then become forgotten. To preserve the algorithms, one needs to constantly copy their source code from old computers to newer ones and preserve the tools needed to compile them.
I plan to introduce one effort to preserve mathematical codes: the histoRicalg project. Its primary objective is to find, document, and preserve the originals of the numerical codes that have found their way into the R statistical programming system. In particular, I will summarize my experience with attempts to run the originals of these codes on a modern Linux PC.
Speakers: Arpad Laszlo Lukacs